Your ICP Is Not About Who’s Ready to Buy

Every vendor who sets up haggl eventually hits the same moment: the ICP configuration screen. Define your ideal customer. Describe the buyer worth investing in.

And almost every vendor fills it in the same way. They describe a customer who is ready to buy. Actively evaluating. Dissatisfied with their current provider. Price-sensitive enough to respond to a deal. Showing purchase intent signals.

This is wrong. Not slightly off — structurally wrong. And it leads to a discount strategy that costs money without generating the returns that would justify it.

The ICP in agentic commerce isn’t about who is likely to convert. It’s about who is worth discounting to acquire. Those are completely different questions — and conflating them is the most expensive mistake a vendor can make when configuring their negotiation logic.

Why Conversion Signals Don’t Belong in Your ICP

In traditional marketing, the ICP exists to focus acquisition effort. You target the people most likely to buy, because reaching everyone is expensive and attention is scarce. Intent signals — “actively shopping,” “dissatisfied with current provider,” “evaluated three competitors” — help you find the hot leads in a cold pool.

Agentic commerce has already done that work. By the time an AI agent negotiates on your site, it is carrying a qualified buyer. The user delegated their shopping to the agent precisely because they have a real need. The agent visited your site because you made the candidate list. You are not fighting for attention. You are at the table.

Conversion — the decision to buy — is what the price negotiation handles. The whole point of the protocol is that a personalized offer, calibrated to the buyer’s profile, closes the deal better than a static list price. You don’t need the ICP to identify “intent.” Intent is already present. The agent is there.

What the ICP actually needs to answer is a different question: given that this buyer converts, how much is that conversion worth over time? That’s the number that determines how deep a discount is rational to offer.

The ICP as a Discount Authorization Framework

Think of the ICP not as a targeting tool but as a financial decision rule. It answers: for this customer profile, what is the maximum discount I should authorize?

The logic is straightforward. A customer who will stay for four years and expand their usage over time is worth more than a customer who churns after three months. The first customer will generate more lifetime revenue — so you can afford to discount more to acquire them. The second customer might convert at the same price, but the economics of their lifetime value don’t justify the same investment.

If your ICP includes intent signals, you will discount customers who happen to be shopping right now — without knowing whether they’ll still be customers a year from now. You’re optimizing for the conversion event, not for what comes after it.

If your ICP includes LTV signals, you will discount customers who are likely to generate high lifetime revenue — and the conversion follows naturally, because you can afford to make a compelling offer to the people who are worth it.

The Four Dimensions of LTV Your ICP Should Capture

Lifetime value is not a single number. It’s shaped by four distinct forces, each of which can be approximated from signals an AI agent can surface about the buyer it’s representing.

1. Retention

How long will this customer stay? The strongest retention predictors are structural, not attitudinal. Ownership depth (do they use the product daily or occasionally?), switching cost (is the product embedded in their workflow?), and dependency (does their team rely on it?). A customer who integrates your product deeply into operations is expensive to replace — they stay. A customer who uses you peripherally churns when a competitor offers 10% off.

Attitudinal signals — “satisfied with their current situation,” “not actively looking” — are worthless retention predictors. People who are “not actively looking” switch all the time. What keeps customers is cost of exit, not sentiment.

2. Expansion

Will this customer spend more over time? Multi-site operators, growing teams, companies in scale phases, buyers who are currently under-using the product category — these are expansion candidates. A company adding 20 employees a quarter is a different customer than a company that’s been at 15 people for three years. One grows into your higher tiers. One doesn’t.

Expansion signals are often structural: headcount trajectory, number of locations, usage maturity in adjacent product categories. Agents can surface many of these from publicly available data.

3. Cost Efficiency

How expensive is this customer to serve? A digitally mature buyer who self-serves, pays reliably, and rarely opens support tickets is worth more per dollar of revenue than a buyer who needs hand-holding through every invoice. Two customers paying identical amounts can have very different actual margins if one consumes ten times the support resources.

Cost-efficiency signals include payment reliability, digital maturity, organizational complexity, and support surface area. A procurement-heavy enterprise buyer who routes every question through a committee costs more to serve than a founder who self-serves.

4. Network Value

Will this customer generate referrals, case studies, or category influence? Investors, accelerator operators, community builders, and industry analysts carry disproportionate network value. A single high-influence customer who recommends you to their portfolio or community can generate ten downstream customers. That multiplier belongs in the LTV calculation even if it doesn’t show up in their direct account revenue.

This is also where “price sensitivity” as a signal reveals its hidden cost. A price-sensitive customer is not a high-LTV customer — they’re a churn risk. The next time a competitor undercuts you, they leave. Including price sensitivity as an ICP positive is literally selecting for the customers most likely to abandon you.

What This Looks Like in Practice

A SaaS vendor building their ICP for haggl might be tempted to write: “companies actively evaluating project management software, currently on a competitor, frustrated with their current tool.” This is a classic intent-based ICP. It will find people who are ready to convert. It will not find people worth converting.

The same vendor, thinking in LTV terms, writes something different: “companies with 10–50 employees in a growth phase (headcount up 20%+ over 12 months), cross-functional teams where project coordination is a daily workflow, with at least two departments already using SaaS tools for operations.”

The second ICP describes customers who will use the product deeply (retention), grow into higher tiers (expansion), are digitally native (cost efficiency), and operate in the kind of network where a recommendation travels (network value). The agent is instructed to surface signals confirming this profile.

A company that matches the first ICP gets a modest discount — converting them isn’t hard, but it’s not clear they’ll stick around. A company that matches the second ICP gets a deep discount, because the math of their expected lifetime value makes the investment obvious.

The Signals That Actually Matter

An AI agent operating on behalf of a buyer can surface a surprising amount of LTV-relevant information: team size and trajectory, industry, technology stack, operational complexity, whether the company is in growth or maintenance mode. The negotiation protocol can request specific proof of match — and the discount awarded can scale continuously with how well the buyer’s profile matches the LTV criteria.

This is very different from asking the agent to prove intent. Intent is binary and ephemeral — the buyer either is or isn’t shopping right now. LTV is dimensional and durable — the signals that predict high lifetime value don’t change based on where the buyer is in their buying cycle.

The practical effect: your highest discounts go to the customers who will generate the most revenue over time, not to whoever happened to be shopping most urgently this week. The discount budget is deployed where it has the highest return, not where the conversion probability is highest.

One Test to Apply to Your ICP

Before you finalize any ICP segment, apply this test: if a customer perfectly matches this profile but churns after 60 days, does the discount you offered still make economic sense?

If the answer is yes, your ICP is probably capturing something real about long-term value. If the answer is no — if the only thing justifying the discount is the assumption that the customer stays — then you have intent signals masquerading as LTV signals. The customer looked like a great acquisition. They just didn’t stay long enough to be one.

Build your ICP around customers who are worth discounting even if they churn at average rates. The customers who match that profile and also stay longer than average are your real payoff.

Start With the Right Question

When you sit down to configure your ICP, the question is not: “Who is most likely to buy from me right now?” The negotiation handles that. The question is: “If I could pick any customer to win today, which one would I most want to still have as a customer in three years?”

Describe that customer. Describe what makes them structurally valuable — not what makes them actively shopping. Then set a discount ceiling that reflects how much you’d pay to acquire them if you knew for certain they’d stay.

That’s the ICP. Everything else is just intent data — and the agent is already carrying that for free.

• Get your embed code — Configure your ICP and start capturing agent traffic today.
• Agents already visit your site — Why you can’t see them, and how to intercept them before they recommend someone else.
• Cove Coffee case study — How LTV-based ICP segments drove a 90% agent recommendation rate.
• Protocol docs — Full technical spec, including ICP segment configuration.